Safe driving behavior reward and punishment system and method and medium

Abstract

The invention relates to the technical field of automobile driving, in particular to a safe driving behavior reward and punishment system and method and a medium, and the safe driving behavior reward and punishment system comprises a data collection module which is at least used for collecting traffic data in real time; the feature extraction module is at least used for performing driving behavior data extraction on the acquired traffic data based on a preset driving behavior extraction rule, and converting the extracted driving behavior data into text description; the preset model module is at least used for generating context embedding with semantic features and grammar features based on the semantic features and grammar information in the text description; the driving behavior evaluation module is at least used for processing context embedding through the attention mechanism and the relation classifier, generating sentence-level driving behavior representation and generating a driving behavior evaluation result based on the driving behavior representation, and the problems that driving behaviors are not fed back in time and driver behaviors are difficult to standardize are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of automotive driving, and particularly relates to a safety driving behavior reward and punishment system, method and medium. Background Art

[0002] With the continuous development of intelligent transportation systems, the accurate assessment of driving behavior has become the key to improving traffic safety. However, existing driving behavior assessment methods have some limitations. Traditional assessment methods may rely on manual observation or simple automated detection, which often have problems such as strong subjectivity and low accuracy. In addition, many existing traffic safety systems cannot provide real-time feedback on driving behavior, making it difficult for drivers to adjust their behavior in a timely manner.

[0003] With the rapid increase in the number of vehicles, traffic safety issues have become increasingly severe, especially in some special traffic scenarios, such as traffic problems at intersections. Accurately assessing driving behavior and promptly giving rewards and punishments are crucial for enhancing drivers' safety awareness and standardizing driving behavior. However, in existing technologies, the reward and punishment measures for driving behavior are often not scientific and systematic enough, lacking clear assessment criteria and automated reward and punishment decision-making mechanisms, resulting in the fairness and effectiveness of rewards and punishments being affected. These problems are particularly prominent in complex traffic environments such as intersections. Summary of the Invention

[0004] In view of the above-mentioned disadvantages of the prior art, the purpose of the present invention is to provide a safety driving behavior reward and punishment system, method and medium, which solves the problems of untimely driving behavior feedback and difficulty in standardizing drivers' behavior.

[0005] To achieve the above object, the present invention adopts the following technical solutions.

[0006] The first aspect of the present invention provides a safety driving behavior reward and punishment system, including: A data acquisition module, at least used for real-time acquisition of traffic data; A feature extraction module, at least used for extracting driving behavior data from the acquired traffic data based on preset driving behavior extraction rules, and converting the extracted driving behavior data into a text description; A preset model module, at least used for generating a context embedding with semantic features and syntactic features based on the semantic features and syntactic information in the text description; A driving behavior assessment module, at least used for processing the context embedding through an attention mechanism and a relation classifier to generate a sentence-level driving behavior representation, and generating a driving behavior assessment result based on the driving behavior representation.

[0007] As an alternative embodiment, the safe driving behavior reward and punishment system further includes a reward and punishment decision-making module, and the reward and punishment decision-making module is at least used to generate a reward and punishment decision based on the driving behavior evaluation result.

[0008] As an alternative embodiment, the safe driving behavior reward and punishment system further includes a feedback module, and the feedback module is at least used to enable the driving behavior evaluation module to feedback the driving behavior evaluation result and / or enable the reward and punishment decision-making module to feedback the reward and punishment decision to the corresponding driver through the feedback module.

[0009] As an alternative embodiment, the reward and punishment decision-making module further includes a reward and punishment determination module and a points management module; wherein: The reward and punishment determination module is at least used to determine whether to give a driver a reward or punishment based on the driving behavior evaluation result; the points management module is at least used to record the situation of the driver's reward or punishment according to a preset points setting rule.

[0010] The second aspect of the present invention provides a method for rewarding and punishing safe driving behaviors, including: Collect traffic data in real time; Extract driving behavior data from the collected traffic data based on a preset driving behavior extraction rule, and convert the extracted driving behavior data into a text description; Based on the semantic features and grammatical information in the text description, generate a context embedding with semantic features and grammatical features; process the context embedding through an attention mechanism and a relation classifier to generate a sentence-level driving behavior representation, and generate a driving behavior evaluation result based on the driving behavior representation.

[0011] As an alternative embodiment, after generating the driving behavior evaluation result based on the driving behavior representation, it includes: Generate a reward and punishment decision based on the driving behavior evaluation result; Feedback the driving behavior evaluation result and / or the reward and punishment decision to the corresponding driver.

[0012] As an alternative embodiment, the traffic data at least includes vehicle position, vehicle driving trajectory, vehicle speed, vehicle acceleration, road speed limit, traffic signal status, and weather condition.

[0013] As an alternative embodiment, the preset driving behavior extraction rule at least includes: whether the vehicle runs a red light, whether the vehicle speeds, and whether the vehicle yields to pedestrians.

[0014] The third aspect of the present invention provides an electronic device, including: At least one processor; and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor can execute the steps of the method described in the first aspect of the present invention by invoking the program instructions.

[0015] The fourth aspect of the present invention provides a readable storage medium storing a computer program, and the computer program is executed by a processor to perform the steps of the method described in the first aspect of the present invention. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a block diagram of a safe driving behavior reward and punishment system according to a specific embodiment of the present invention.

[0018] Figure 2 It is a schematic flowchart of a safe driving behavior reward and punishment system according to a specific embodiment of the present invention.

[0019] Figure 3 It is a schematic structural diagram of an electronic device according to an embodiment of the present invention. Detailed Embodiments

[0020] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of the present application. In addition, it should be understood that the specific embodiments described here are only used to illustrate and explain the present application, and are not used to limit the present application.

[0021] It should be noted that the description order of the following embodiments does not limit the preferred order of the embodiments of the present application. And in the following embodiments, each embodiment is described with its own emphasis. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0022] In recent years, some advanced technologies in the field of natural language processing, such as the Bidirectional Encoder Representations from Transformers (BERT) model based on the self-attention mechanism, have demonstrated excellent performance in text understanding and processing, which provides new ideas for driving behavior evaluation. However, directly applying the BERT model to driving behavior evaluation still poses challenges. For example, the standard BERT model may not fully adapt to the characteristics of traffic data and needs to be further optimized to improve the accuracy of evaluation.

[0023] As Figure 1 shown in the figure, the first aspect of the present invention provides a reward and punishment system for safe driving behavior, including a data acquisition module, a data preprocessing module, a feature extraction module, a preset model module, a driving behavior evaluation module, a reward and punishment decision-making module, a feedback module, and a self-learning optimization module.

[0024] The data acquisition module is at least used to collect traffic data in real time.

[0025] Specifically, taking the intersection scenario as an example, the data acquisition module collects vehicle position, vehicle driving trajectory, speed, acceleration, traffic signal status, road speed limit, vehicle position, pedestrian position, and weather conditions in real time through means such as cameras, sensors, V2X, and vehicle networking.

[0026] The data preprocessing module is at least used to preprocess the original traffic data obtained by the data acquisition module, including but not limited to data cleaning, data standardization, data completion, data encryption and anonymization, and data compression, etc. These preprocessing steps improve the data quality and usability.

[0027] The feature extraction module is at least used to extract driving behavior data based on preset driving behavior extraction rules from the collected traffic data or the traffic data preprocessed by the data preprocessing module, and convert the extracted driving behavior data into text descriptions.

[0028] Specifically, in this system, the preset driving behavior extraction rules are a set of predefined standards or conditions for identifying and extracting specific driving behavior data from traffic data. These preset driving behavior extraction rules provide a framework for this system.

[0029] Exemplarily, the preset driving behavior extraction rules include: 1. Whether the vehicle runs a red light: Compare the vehicle position in the traffic data with the traffic signal status. 2. Whether the vehicle speeds: Compare the vehicle speed in the traffic data with the road speed limit. 3. Whether the vehicle yields to pedestrians: Analyze the data of the vehicle position and the pedestrian position.

[0030] These rules provide a extraction standard for the system to identify and extract specific driving behavior data from complex traffic data.

[0031] Specifically, the text description is to convert the extracted driving behavior data into a language form readable by a preset model, which is the input of the preset model. For example: the text description is: The vehicle passes through the intersection when the traffic light is red; The vehicle actively gives way to pedestrians; For giving way to pedestrians: "Vehicle C stops in front of the crosswalk to let pedestrians pass".

[0032] The preset model module, including a contraction path and an expansion path, is at least used to generate a context embedding with semantic features and syntactic features based on the semantic features and syntactic information in the text description.

[0033] The driving behavior evaluation module is at least used to process the context embedding through an attention mechanism and a relation classifier to generate a sentence-level driving behavior representation, and generate a driving behavior evaluation result based on the driving behavior representation. Among them, the driving behavior evaluation module outputs an evaluation result of the driving behavior through the processing and calculation of the BERT model.

[0034] In an embodiment of the present invention, the preset model used is specifically an optimized model of the Bidirectional Encoder Representations from Transformers (BERT) of the self-attention model. This model is used to accurately evaluate driving behaviors. The optimization measures of this model include model compression, enhancement of long text processing capabilities, incorporation of external knowledge such as traffic rules, and improvement of pre-training tasks.

[0035] Specifically, model compression specifically includes: reducing the number of parameters and computational complexity of the BERT model through technical means such as pruning and quantization, improving the processing speed, and reducing the consumption of hardware resources.

[0036] Specifically, the enhancement of long text processing capabilities specifically includes: introducing Transformer-XL or other related technologies to enable the BERT model to better process long texts to adapt to complex driving scenario descriptions.

[0037] Specifically, the incorporation of external knowledge such as traffic rules specifically includes: incorporating information such as traffic rules and road signs into the BERT model in a specific way to enhance its understanding ability of driving scenarios and ensure the accuracy of evaluation.

[0038] Specifically, the improvement of the pre-training task specifically includes: on the basis of the traditional Masked Language Model (MLM) and Next Sentence Prediction (NSP), adding pre-training tasks such as contrastive learning and sentence reordering to enrich the pre-training process of the model and improve its generalization ability.

[0039] The optimized BERT model can more accurately understand the semantics described by these texts, so as to conduct a more precise evaluation of driving behaviors. Exemplarily, the optimized BERT model will output an evaluation score indicating the quality of driving behaviors. In addition, the optimized BERT model can process data faster, providing real-time behavior feedback to the driver, which helps the driver adjust the driving manner in a timely manner. The present invention improves the traffic safety level by collecting traffic data in real time, using the optimized BERT model to evaluate driving behaviors, and combining a reward and punishment mechanism to encourage the driver to maintain good driving habits.

[0040] Specifically, the preset model module of the present invention belongs to the front part of the BERT model. In this BERT model, the text description is the input, and the sentence pair is the basic data unit for processing. The sentence pair is extracted from the text description and is directly related to the driving behavior. Such text description makes complex data easier to be understood and analyzed by the preset model, providing a clear basis for subsequent driving behavior evaluation and decision-making.

[0041] Specifically, context embedding refers to embedding the position and relevant information of the current data point (a certain word in the text) in the entire data sequence (the entire sentence or paragraph) into a vector representation. In the optimized BERT model, context embedding not only contains the semantic information of the vocabulary, but also contains the grammatical structure and position information of the vocabulary in the sentence. Context embedding refers to embedding the position and relevant information of the current data point (such as a certain word or phrase in the text) in the entire data sequence (such as the entire sentence or paragraph) into a vector representation. This method can capture the context information in the data and improve the performance of the model.

[0042] In the safe driving scenario at an intersection, when processing the phrase "pedestrians on the zebra crossing", the present invention can utilize the context information of the previous words "pedestrians", "on", and "on the zebra crossing", and embed this information into the vector representation, so as to more accurately predict that the next word may be "cross the road" rather than other irrelevant words.

[0043] Semantic features and grammatical information are a kind of representation inside the model and are used for the subsequent processing of the model. These two should not be considered as the intermediate outputs of the preset model module. They are just formed when the model processes the input data. Context embedding is a vector representation generated by the model when processing the input data, and it contains the context information of the input data.

[0044] The relationship between context embeddings with semantic features and syntactic features is complementary. Semantic features provide the true meaning and intention of words or sentences, while syntactic information provides the structural and positional relationships of words or sentences within a sentence. These two types of features together form the basis of context embeddings, enabling the present invention to more accurately understand and process information in text descriptions.

[0045] For example: When processing the text description "There are pedestrians crossing the road ahead", a context embedding vector with semantic and syntactic features may be generated. This vector not only contains information about the two key semantic features "pedestrians" and "crossing the road", but also information about their syntactic structure and positional relationship. In this way, it is possible to more accurately judge the safety of the current road conditions and make corresponding reward and punishment decisions.

[0046] In the scenario of safe driving at intersections, for the traffic rule "Stop at red lights and go at green lights", its syntactic information includes the association between "red lights" and "stop", and the association between "green lights" and "go". These syntactic rules help the system understand the meaning of traffic signals and thus guide driving behavior.

[0047] In text processing, the contracting path and the expanding path generally refer to two stages in the BERT model (such as the U-Net or variant networks), and they play different roles in processing and generating context embeddings.

[0048] The main function of the contracting path is to extract high-level features of the input text. Through layer-by-layer convolution and pooling operations, the spatial dimension of the feature map is gradually reduced while increasing the degree of abstraction of the features. In the contracting path, the model gradually encodes the input text, transforming the original information of the text into a more compact feature representation. This process is similar to a traditional encoder, compressing the input data into a low-dimensional feature space. The feature representation generated by the contracting path contains the semantic features and syntactic information of the text, but these features are spatially compressed.

[0049] The main function of the expansion path is to gradually restore the high-level features extracted in the contraction path to the original spatial dimension. Through layer-by-layer deconvolution (or upsampling) operations, the spatial dimension of the feature map is gradually increased while maintaining the level of feature abstraction. In the expansion path, the model gradually decodes the compressed feature representation and restores it to the same spatial dimension as the original input. This process is similar to a traditional decoder, which restores a low-dimensional feature representation to a high-dimensional feature map. The feature representation generated by the expansion path not only contains the semantic features and syntactic information of the text but also restores the spatial structure of the features, enabling the model to more accurately understand and process the information in the text description.

[0050] The common goal of the contraction path and the expansion path is to generate context embeddings with semantic and syntactic features, enabling the model to more accurately understand and process the information in the text description. Exemplarily, when processing the text description "There is a pedestrian crossing the road ahead", the contraction path extracts key semantic features such as "pedestrian" and "crossing the road" and compresses these features. The expansion path then restores these compressed features to the original spatial dimension while retaining information about the syntactic structure and positional relationships. The resulting context embedding vector can more accurately reflect the meaning and structure of the text, helping the system to judge the safety of the current road conditions. In this way, the contraction path and the expansion path work together to enable the model to better understand and process complex text information.

[0051] Exemplarily, the driving behavior represents an evaluation score. The attention mechanism is one of the core components in the BERT model and is used to enable the model to dynamically focus on different parts of the input sequence when processing it. In the driving behavior reward and punishment system, the attention mechanism is used in the following three aspects: Feature selection: The model can automatically select the features that are most important for the evaluation of driving behavior, giving them higher weights. The weights of safety distance, driving speed, and traffic signal compliance will change dynamically and be adaptively adjusted according to the driver's behavior, with the sum of the weights being 1. Context understanding: Through the attention mechanism, the model can better understand the context environment in which the driving behavior occurs, including road conditions and traffic signals. Behavior pattern recognition: The model can identify patterns in driving behavior, including safety distance, driving speed, and traffic signal compliance. The relation classifier in the BERT model is used to identify and understand the relationships between different elements in the input sequence. In the driving behavior reward and punishment system, the relation classifier is mainly used in the following two aspects: Traffic rule compliance: The model can identify whether the driving behavior complies with traffic rules, i.e., whether running a red light and speeding occur. Vehicle interaction: The model can understand the interaction behavior between different vehicles, i.e., the safety distance.

[0052] It should be noted that the method of the present invention for realizing driving behavior evaluation and rewards and punishments by using an improved BERT model based on actual safe driving scenarios is flexible. The modules listed in the present invention are not strictly divided. That is, the modules provided by the present invention can be combined according to actual situations or further split or named based on technical features, and all belong to the scope protected by the present invention. For example, the preset model module and the driving behavior evaluation module provided by the present invention and other modules all participate in the construction of the BERT model. Therefore, the preset model module and the driving behavior evaluation module can be collectively referred to as a part of the BERT model, or only the driving behavior evaluation module can be regarded as a part of the BERT model, and the preset model module can be regarded as a non-final state or pre-stage of the BERT model, that is, regarded as not belonging to the BERT model.

[0053] In this way, the present invention can accurately evaluate driving behaviors, encourage drivers to actively improve their driving habits and reduce dangerous behaviors. At the same time, it provides valuable data support for traffic management departments, enabling them to formulate more accurate and effective traffic management strategies, which helps to reduce the accident rate, optimize traffic flow, and improve road use efficiency.

[0054] In an embodiment of the present invention, the safe driving behavior rewards and punishments system further includes a rewards and punishments decision-making module, and the rewards and punishments decision-making module is at least used to generate rewards and punishments decisions based on the driving behavior evaluation results.

[0055] Specifically, based on the evaluation results of the preset model module, this module formulates and executes corresponding rewards and punishments measures. According to clear driving behavior evaluation criteria, such as safety distance, driving speed, signal light compliance, etc., it quantitatively scores driving behaviors. According to the scoring results, it gives integral rewards to drivers who show safe driving behaviors; and deducts points for dangerous driving or illegal driving behaviors.

[0056] Here, in this way, it can not only encourage drivers to abide by traffic rules, but also restrain unsafe driving behaviors.

[0057] In an embodiment of the present invention, the rewards and punishments decision-making module further includes a rewards and punishments determination module and an integral management module.

[0058] The rewards and punishments determination module is at least used to determine whether to give rewards or punishments to drivers based on the driving behavior evaluation results. For good driving behaviors such as abiding by traffic rules and giving way to pedestrians, the system will give integral rewards; while for dangerous driving behaviors such as running red lights and speeding, the system will deduct points.

[0059] The integral management module is at least used to record the situations of drivers' rewards or punishments according to preset integral setting rules. The integral can be used to exchange for coupons, reduce traffic violation fines and other benefits, so as to encourage drivers to maintain good driving habits.

[0060] In one embodiment of the present invention, the safe driving behavior reward and punishment system also includes a feedback module, and the feedback module is at least used for the driving behavior evaluation module to feed back the driving behavior evaluation result and / or the reward and punishment decision module to feed back the reward and punishment decision to the corresponding driver through the feedback module.

[0061] Here, through the combination of the reward and punishment determination module and the point management module, driving behavior can be objectively and impartially evaluated, and immediate rewards or punishments can be provided, which further improves the level of traffic safety, encourages drivers to have good driving habits, and can encourage drivers to abide by traffic rules, while restraining unsafe driving behaviors and providing drivers with a clear direction for improvement. The introduction of the feedback module further enhances the practicality of the system, allowing drivers to understand their driving performance and corresponding reward and punishment results in a timely manner, and will also increase the stickiness of users and point service providers.

[0062] The solution provided by the present invention can accurately evaluate driving behavior, provide real-time feedback, and implement a scientific reward and punishment mechanism system to effectively improve the level of traffic safety. It can help drivers adjust their driving behavior in a timely manner and provide valuable data support for traffic management departments, thereby formulating more scientific and reasonable traffic management strategies.

[0063] In one embodiment of the present invention, the feedback module further includes a real-time feedback module and a historical record query module.

[0064] The real-time feedback module is at least used for the system to provide real-time feedback to the driver through the vehicle-mounted device or mobile phone APP, informing him of the current driving behavior evaluation results and rewards and punishments, which helps the driver to adjust his driving behavior in time.

[0065] Taking a vehicle passing through an intersection as an example, the history record query module is used at least for the driver to query his driving behavior history at any time through the vehicle-mounted device, mobile phone APP or other communication methods, including the evaluation score, rewards and punishments each time he passes through the intersection, which helps the driver understand his driving habits and make improvements. At the same time, the module can also provide traffic safety education and training resources to help drivers improve their safety awareness.

[0066] In one embodiment of the present invention, the safe driving behavior reward and punishment system also includes a self-learning optimization module, which is at least used to continuously collect new traffic data and driving behavior evaluation results, and use these data to retrain and optimize the optimized BERT model, thereby maintaining the advancement and accuracy of the model by continuously updating model parameters and introducing new training samples.

[0067] Traditional driving behavior assessment methods may rely on manual observation or simple automated detection, which often suffer from strong subjectivity and low accuracy. By introducing an optimized BERT model, the present invention improves the accuracy and objectivity of driving behavior assessment, enabling more accurate assessment of driving behavior and reducing misjudgment and missed judgment situations.

[0068] Existing traffic safety systems often fail to provide real-time feedback on driving behavior. By collecting and processing traffic data in real time, the present invention can immediately feedback the assessment results to the driver, helping them correct improper driving behavior in a timely manner and enhancing their safety awareness.

[0069] The reward and punishment measures for driving behavior are often not scientific and systematic enough. Through clear assessment criteria and automated reward and punishment decision-making, the present invention makes the rewards and punishments more fair, reasonable, and effective, and can more scientifically encourage safe driving behavior and curb dangerous driving and violations.

[0070] In summary, by introducing an optimized BERT model, a real-time feedback and prompting mechanism, and a scientific reward and punishment decision-making system, the present invention effectively solves the defects in the prior art and brings significant beneficial effects.

[0071] As Figure 2 shown, a method for rewarding and punishing safe driving behavior according to the second aspect of the present invention includes: Step S100: Collect traffic data in real time, where the traffic data at least includes vehicle position, vehicle driving trajectory, vehicle speed, vehicle acceleration, road speed limit, traffic signal status, and weather conditions; Step S200: Extract driving behavior data from the collected traffic data based on preset driving behavior extraction rules, and convert the extracted driving behavior data into text descriptions. The preset driving behavior extraction rules at least include: whether the vehicle runs a red light, whether the vehicle speeds, and whether the vehicle yields to pedestrians; Step S300: Based on the semantic features and syntactic information in the text description, generate a context embedding with semantic features and syntactic features; Step S400: Process the context embedding through an attention mechanism and a relation classifier to generate a sentence-level driving behavior representation, and generate a driving behavior assessment result based on the driving behavior representation.

[0072] In an embodiment of the present invention, after generating a driving behavior assessment result based on the driving behavior representation, it includes: Step S500: Generate a reward and punishment decision based on the driving behavior assessment result; feedback the driving behavior assessment result and / or the reward and punishment decision to the corresponding driver.

[0073] Based on the same idea as the system in the above embodiments, the method provided by the present invention can implement the method of the above embodiments.

[0074] The third aspect of the present invention provides an electronic device, including: At least one processor; and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor can execute the steps of the method according to any one of the above embodiments by invoking the program instructions.

[0075] The fourth aspect of the present invention discloses a readable storage medium storing a computer program, and the computer program is executed by a processor to perform the steps of the method according to any one of the above embodiments.

[0076] The computer-readable storage medium may include: any entity or device capable of carrying the computer program, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), and software distribution medium, etc. The computer program includes computer program code. The computer program code may be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), and software distribution medium, etc.

[0077] Any process or method description represented in the flowchart or described in other ways herein can be understood as representing a module, segment or part of code including one or more executable instructions for implementing a specific logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations, where the functions may be executed in a substantially simultaneous manner or in a reverse order according to the involved functions, rather than in the order shown or discussed, which should be understood by those skilled in the technical field to which the embodiments of the present invention belong.

[0078] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a sequenced list of executable instructions for implementing a logical function, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processing module, or other systems that can fetch and execute instructions from the instruction execution system, apparatus or device), or used in combination with these instruction execution systems, apparatus or devices.

[0079] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A safe driving behavior reward and punishment system, characterized in that: include: A data collection module, at least for collecting traffic data in real time; A feature extraction module, at least for extracting driving behavior data from the collected traffic data based on a preset driving behavior extraction rule, and converting the extracted driving behavior data into a text description; A preset model module is at least used to generate context embedding with semantic features and grammatical features based on semantic features and grammatical information in the text description; The driving behavior evaluation module is at least used to process the context embedding through an attention mechanism and a relation classifier, generate a sentence-level driving behavior representation, and generate a driving behavior evaluation result based on the driving behavior representation.

2. The safe driving behavior reward and punishment system according to claim 1, characterized in that: It also includes a reward and punishment decision module, which is at least used to generate reward and punishment decisions based on the driving behavior evaluation results.

3. The safe driving behavior reward and punishment system according to claim 2, characterized in that: It also includes a feedback module, which is at least used for the driving behavior evaluation module to feed back the driving behavior evaluation result and / or the reward and punishment decision module to feed back the reward and punishment decision to the corresponding driver through the feedback module.

4. The safe driving behavior reward and punishment system according to claim 2, characterized in that: The reward and punishment decision module also includes a reward and punishment determination module and a points management module; wherein: The reward and punishment determination module is at least used to determine whether to give a reward or punishment to the driver based on the driving behavior evaluation result; The points management module is at least used to record the driver's reward or punishment according to the preset points setting rules.

5. A method for rewarding and punishing safe driving behavior, characterized in that: include: Collect traffic data in real time; Extracting driving behavior data from the collected traffic data based on preset driving behavior extraction rules, and converting the extracted driving behavior data into text descriptions; Based on the semantic features and grammatical information in the text description, a context embedding with the semantic features and the grammatical features is generated; The context embedding is processed through the attention mechanism and relation classifier to generate sentence-level driving behavior representation, and the driving behavior evaluation result is generated based on the driving behavior representation.

6. The safe driving behavior reward and punishment method according to claim 5, characterized in that: After generating the driving behavior evaluation result based on the driving behavior representation, including: Generate reward and punishment decisions based on driving behavior assessment results; The driving behavior evaluation result and / or the reward and punishment decision are fed back to the corresponding driver.

7. The safe driving behavior reward and punishment method according to claim 5, characterized in that: The traffic data includes at least vehicle position, vehicle driving trajectory, vehicle speed, vehicle acceleration, road speed limit, traffic light status and weather conditions.

8. The safe driving behavior reward and punishment method according to claim 5, characterized in that: The preset driving behavior extraction rules at least include: whether the vehicle runs a red light, whether the vehicle exceeds the speed limit, and whether the vehicle gives way to pedestrians.

9. An electronic device, characterized in that: include: at least one processor; And at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the steps of the safe driving behavior reward and punishment method as described in any one of claims 5-9.

10. A readable storage medium storing a computer program, characterized in that: The computer program is executed by the processor to perform the steps of the safe driving behavior reward and punishment method as described in any one of claims 5-9.

Images